The localization and adsorption of benzene and propylene in ITQ-1zeolite: grand canonical Monte Carlo simulations
T.J. Hou, L.L. Zhu, Y.Y. Li, X.J. Xu*
College of Chemistry and Molecular Engineering, Peking University, Beijing 100871, People's Republic of China
Received 27 March 2000; accepted 6 April 2000
Abstract
Grand canonical Monte Carlo (GCMC) simulations have been performed to investigate the localization and adsorption
behavior of benzene and propylene, which are involved in the cumene synthesis process, in purely siliceous MWW zeolite
(ITQ-1). From the mass clouds of GCMC simulations, it can be seen that the benzene and propylene molecules show different
localization and adsorption behavior in the zeolite cavities. In the 10-MR channels, both benzene and propylene show high
localization. In the 12-MR supercages, the propylene molecules cannot only almost ®ll all the possible positions in one
supercage, but also can be steadily located in the short 10-MR conducts interconnecting the 12-MR supercages, where the
benzene molecules are adsorbed close to three adsorption sites. By analyzing the location of benzene and propylene in ITQ-1, it
can be deduced that the alkylation of benzene and propylene will happen mainly in 12-MR supercages. Moreover, a series of
simulations have been performed to predict the adsorption isotherms of benzene and propylene at 315 K and 0±1.4 kPa. The
results for benzene generally are in agreement with the trend from experiments on a series of aromatic compounds. The results
reveal that at low pressures, the loading of propylene is lower than that of benzene, which seems to be caused by the relatively
unfavorable potential interactions between propylene/zeolite and propylene/propylene. q 2001 Elsevier Science B.V. All
rights reserved.
Keywords: ITQ-1; Grand canonical Monte Carlo simulations; Adsorption; Benzene; Propylene
1. Introduction
The alkylation of benzene with propylene to
produce cumene, a starting material for the production
of acetone and phenol [1], is very important in hydro-
carbon processes. Traditionally, some mineral acids
including AlCl3 and others are employed, which
present good catalytic performance [2,3], while rais-
ing serious environmental problems such as corrosion
and waste disposal. In order to overcome some draw-
backs of the traditional mineral acid catalysts, tech-
nologies such as the Mobil-Badger process using H-
ZSM-5 [4], the CDTEC and ENI process using Y and
b zeolites [5] and the DOW process using Mordenite
[6] have made great progress in recent years. The
zeolites are considered to be cleaner catalysts; more-
over, they can effectively reduce the amount of less
desired products such as diisopropylbenzenes, but
introduce the formation of n-propylbenzene which is
not formed in signi®cant amounts when using mineral
acid catalysts [7].
It is well known that in the case of zeolites, the pore
dimensions and the diffusion of the reactants and
products signi®cantly contribute to reactions that
occurred in zeolites, since the reactivity and ®nal
Journal of Molecular Structure (Theochem) 535 (2001) 9±23
0166-1280/01/$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.
PII: S0166-1280(00)00520-0
www.elsevier.nl/locate/theochem
* Corresponding author.
E-mail address: [email protected] (X.J. Xu).
product distributions will depend highly on the zeolite
channel structure and the diffusion of the sorbate
molecules. For instance, the formation of n-propylto-
luene has been observed with ZSM-5 but not with
Mordenite, which suggests that the transalkylation
reaction occurs in the 10-MR channel intersection of
ZSM-5, whereas the intersection of 12-MR and 8-MR
channels in Mordenite does not provide suf®cient
space for the bimolecular transalkylation process to
occur. However, the yield of transalkylation in zeolite
Y is found to be much lower than that in ZSM-5 under
the same reaction conditions, which suggests that the
reactivity cannot be determined only by the size of
cavities in which the reactions take place, but the
diffusion of the chemical compounds involved in the
reaction also needs to be taken into account. With
respect to energy, the energetic complementarity
between the zeolite and the chemical compounds
involved in the reaction is simply expressed as
shape adsorption or shape selectivity of the zeolite.
The diffusion of the adsorbents in the zeolite is really
a ªdockingº process, and they always prefer to follow
an energetically favorable pathway.
Several zeolites including H-ZSM-5, USY, b and
MCM-22 have been tested for the alkylation of
benzene with propylene, and it has been found that
MCM-22 is very reactive [8]. Zeolite MCM-22 (IZA
code MWW) is a novel zeolite discovered recently by
scientists at Mobil [9,10]. Compared with other
common types of zeolites, MCM-22 possesses an
interesting and unusual framework structure: two
independent pore systems formed by interconnected
sinusoidal 10-MR pores with a 4±5.5 AÊ diameter and
an independent 12-MR supercage with 18.2±7.1 AÊ
linked by 10-MR windows. The unusual framework
topology, high thermal stability, large surface area
and good adsorption capacity render this zeolite
very interesting for catalysis. Several previous studies
have looked into the diffusivities of benzene and
cumene in the MWW structure [11,12]. It has been
found that the diffusion and adsorption behaviors of
these molecules are different in the two independent
pore systems. The simulations of Perego et al. [11]
indicated that the diffusion of cumene was more
favorable in the 12-MR supercages than in the 10-
MR sinusoidal system, although the 10-MR windows
interconnecting the 12-MR supercages still presented
a diffusion barrier for cumene. The molecular
dynamics simulations of Sastre et al. [12] revealed
that benzene was not seen to diffuse through two
pore systems, and only intracage mobility was seen
in the supercage voids where a certain activation
energy must be necessary for a benzene molecule
migrating between 12-MR supercages through 10-
MR windows.
Until now, few studies have been performed to
investigate the localization and adsorption of benzene
and propylene in MCM-22, and the favorable adsorp-
tion sites of those molecules in zeolite lattices are not
clear. As far as we know, no work has been under-
taken yet to investigate the adsorption properties of
zeolite and propylene in MWW type zeolites through
grand canonical Monte Carlo simulations (GCMC)
simulations. In the present work, the GCMC simula-
tion technique was used to predict the adsorption char-
acteristics of benzene and propylene in the ITQ-1
zeolite. We intended to determine the potential
adsorption sites of benzene and propylene in ITQ-1
lattice, as well as the properties of the localization of
benzene and propylene in the zeolite. In the mean-
time, we wanted to predict the adsorption isotherms
for these two kinds of molecules.
2. Method
2.1. Model representations and potential force®eld
Considering the high Si/Al ratio of the MCM-22
type zeolite and the dif®culty of determining Al distri-
butions in disordered zeolites by experiments, mean-
while, in order to simplify the simulations, the pure
siliceous analogue of MCM-22, ITQ-1, was adopted
in this paper. The model of the ITQ-1 was constructed
according to the results from Camblor et al. [13]. In
the simulations, the silicon and oxygen atoms of the
zeolite framework were assumed to be ®xed at their
crystallographic positions from X-ray diffraction
studies. The benzene and propylene molecules were
rigid, that is to say, they could only translate and
rotate, but were not allowed to deform.
The zeolite and the sorbates were assumed to
interact via a pairwise-additive potential between
atoms of the guest molecules and atoms of the
zeolite. The site±site interactions are models with a
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±2310
Lennard-Jones plus Coulomb potential,
V�rij� � Dij
�R0�ijRij
" #12
22�R0�ij
Rij
" #6( )1
qiqj
Rij
�1�
where i and j indicate atoms of the sorbate and of the
zeolite, respectively, and Rij is the distance between
them. Dij and �R0�ij are the Lennard-Jones parameters,
and qi and qj are the partial charge of the atoms. Six
different atom types were considered in the studied
system, including O_z (O atom in zeolite framework),
Si (Si atom in zeolite framework), C_R (C atom in
benzene ring), C_3 (sp3 C in propylene), C_2 (sp2 C in
propylene) and H (H atom in benzene and propylene).
The Lennard-Jones parameters for O_z and Si origin-
ally derived by Burchart [14], and those for other atom
types taken from Mayo [15], are listed in Table 1.
Then, the off-diagonal van der Waals parameters for
each pair of atoms were calculated based on the
geometric mean. The partial charges for O_z
(20.19ueu) and Si (10.38ueu) were taken from the
calculations of Burchart. The above hybrid force®eld
has been tested by MSI and distributed as a Burchart±
Dreiding force ®eld [16]. The partial charges for the
atoms in benzene and propylene (Fig. 1) were
computed using the AM1 method, available in
MOPAC 7.0 [17].
2.2. Grand canonical Monte Carlo simulations
The GCMC simulation may be the most common
technique for predicting the zeolite adsorption phase
equilibria from molecular simulations [18±20]. The
GCMC simulation technique simulates the equili-
brium of a collection of adsorbates in a micropore at
constant chemical potential, volume, and temperature
or pressure. In the GCMC simulations, the number of
particles in the system is not ®xed, but the chemical
potential of each species is ®xed. Sorption translates
the chemical potential into the partial pressure (or
fugacity) of each component. Equilibrium is achieved
when the temperature (Tframe) and the chemical poten-
tial (m frame) of the gas inside the framework are equal
to the temperature (Tgas) and chemical potential (m gas)
of the free gas outside the framework. For a non-ideal
gas, the chemical potential depends upon fugacity ( f ),
which is a function of both temperature and pressure.
The bulk pressure can be determined from the chemi-
cal potential using a Lennard-Jones equation. So the
GCMC simulation technique enables one to study
many important characteristics of zeolite systems
under certain pressure and temperature.
Eight unit cells of zeolite with a total of 1728 atoms
were used to construct the simulation box, and peri-
odic boundary conditions were applied in three
dimensions in order to simulate an in®nite (macro-
scopic) system. In order to achieve the real equilibra-
tion of the system, the length of the simulations was
totally 6 £ 106 steps, and every 600 steps, a con®gura-
tion of the system was recorded. The ®rst three million
steps were used for equilibration and not included in
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±23 11
Table 1
Lennard-Jones parameter for six types of atoms
Atom type D0 (kcal mol21) R0 (AÊ )
O_z 0.1648 3.3000
Si_z 0.0496 4.2000
C_R 0.0951 3.8983
C_3 0.0951 3.8983
C_2 0.0951 3.8983
H 0.0152 3.1950
Fig. 1. Partial charges for: (a) benzene; and (b) propylene used in calculating the zeolite/sorbates and sorbates/sorbates interaction energy, given
in units of the electron charges ueu.
T.J.
Ho
uet
al.
/Jo
urn
al
of
Mo
lecula
rStru
cture
(Theo
chem
)535
(2001)
9±
23
12
Fig. 2. Skeletal drawings of the framework structure of ITQ-1 (MWW structure type). (a) Schematic view of the independent pore systems along the yz axis. (b) Schematic view of
the independent pore systems along the xy axis. (c) Two adjacent 12-MR supercages. The sinusoidal 10-MR channels are all interconnected to each other, and multiple diffusion
trajectories can be allowed to every diffusing molecule. The 12-MR are independent with 10-MR channels, which are interconnected through short 10-MR windows.
the averaging. A cutoff of 10 AÊ was applied to the
Lennard-Jones interactions, and the long-range elec-
trostatic interactions were calculated by using the
Ewald summation technique. The Ewald summation
to calculate the adsorbate±adsorbent and adsorbate±
adsorbate is generally time-consuming. We therefore
used a grid-interpolation procedure in which the simu-
lation boxes were split into a collection of small
cubes. The grid-interpolation method allows us to
take into account any degree of accuracy in the
description of the adsorbate/zeolite and adsorbate/
adsorbate interaction energy since all the needed
grids are calculated separately prior to any simulation
runs. First, the GCMC simulations were carried out in
the condition of 300 K and 1 atm. Then, a series of
simulations were performed to predict the adsorption
isotherms for benzene and propylene at 315 K. All
calculations were performed in the Cerius2 molecular
simulation package [16] on a SGI Octane 2-CPU
workstation.
In this paper, we also computed the potential
energy of the diffusing benzene molecules along
their trajectory, which allowed estimation of the
energy pro®les of one benzene molecule while it
diffused between two adjacent 12-MR supercages
through 10-MR windows in the ITQ-1 zeolite. Conse-
quently, the activation energy of one benzene mole-
cule through 10-MR windows was predicted.
3. Results and discussion
A deep insight into the channel systems present in
ITQ-1 (Fig. 2) reveals some special features that will
greatly affect the location and adsorption of sorbates
in the zeolite lattice. First, all 10-MR sinusoidal chan-
nels are interconnected with each other and have high
tortuosity. The benzene molecules and even the
propylene ones may be restricted through sinusoidal
channels in ITQ-1. Second, the larger 12-MR cavities
have large dimensions with 7.1 AÊ £ 18.2 AÊ , so they
are expected to host both benzene and propylene
whose mobility and location inside the supercages
will be very interesting to investigate. Additionally,
we are also interested in knowing whether they
migrate from one cavity to another nearby through
10-MR openings or tend to remain inside a given
cavity. Certainly, all these features will depend on
the conditions of the environment including tempera-
ture and pressure.
3.1. The in¯uence of the minor distortions of
framework to simulation results
In the calculations, the zeolite atoms were ®xed at
their crystallographic positions. Generally, the frame-
work of zeolite is relatively rigid and possesses good
thermal stability. For instance, the introduction of
xylene into the cavities of the faujasite NaY has little
effect on the framework structure [21]. In order to
examine the in¯uence of the minor change of zeolite
lattice on the simulation results, the framework of the
ITQ-1 zeolite was manually adjusted to obtain three
transformative structures: the ®rst structure was
obtained by expanding (by averages) the coordinates
of all atoms by 1% along the a and b directions; the
second one was obtained by expanding (by averages)
the coordinates by 1.5% along the a and b directions;
and the third one was obtained by expanding (by
averages) the coordinates by 2% along the c direction.
Compared with the volume of the crystallographic
structure, the volumes of these three adjusted
structures are 1, 2.75 and 2% larger (the cell para-
meters for these three adjusted models are listed in
Table 2). For these three adjusted structures and the
crystallographic structure, the simulations of benzene
in ITQ-1 were performed by using GCMC simulations
at standard temperature and pressure.
The full loadings of benzene and the interaction
energy for most probable distributions (Table 2) indi-
cate that these four models do not show noticeable
differences. Moreover, four mass clouds for benzene
molecules are similar, indicating that the minor
distortion of the zeolite framework does not have
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±23 13
Table 2
The cell parameters and the GCMC simulation results for four
models
Model a (AÊ ) b (AÊ ) c (AÊ ) Loading E (kJ mol21)a
1 14.2081 14.2081 24.9452 59.71 281.9
2 14.3502 14.3502 24.9452 58.14 281.1
3 14.4212 14.4212 24.9452 62.55 282.3
4 14.2081 14.2081 25.4410 60.11 280.7
a The average interaction energy between benzene and the zeolite
framework.
considerable effect on the simulation results. So in this
paper, the crystallographic structure of ITQ-1 is
adopted and the zeolite ¯exibility is ignored.
3.2. The diffusions of benzene in the two independent
channel systems
The evolution of the loading and energy of benzene
shows that after 6 £ 106 simulation steps, the equili-
bration has been achieved. At 300 K and 1 atm, the
full loading of benzene is about 63 molecules per 8
ITQ-1 unit cells. Fig. 3(a) depicts the energy distribu-
tion of interaction between benzene and ITQ-1
zeolite at full loadings. The energy distribution
is roughly single-peaked, with a maximum
around 219.9 kJ mol21, and a shoulder from
217.5±215.0 kJ mol21.
In order to characterize the location of the adsorbed
molecules in the ITQ-1 zeolite, several mass clouds
were depicted. As a powerful analysis tool, the mass
cloud shows the preferred positions of the sorbates in
the zeolite. The mass cloud of benzene, with respect to
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±2314
Fig. 3. (a) Benzene/zeolite potential energy distribution. (b) Propylene/benzene potential energy distribution.
the zeolite famework, is shown in Fig. 4. The center of
mass for each sorbate molecule in each con®guration
is displayed as a dot in the model space. From Fig.
4(a), it can be noted that the spatial distribution of
benzene is roughly territorial, which can be divided
into four regions: one in the 10-MR channels and
three others in the 12-MR supercages, which have
been named as S1, S2, S3 and S4. The S1 site is
located in the 10-MR channels. The S2 site is
observed near the 10-MR facing the 6-MR in
supercage, while both the S3 and S4 sites are close
to the central part of the 12-MR supercages. The
distribution of benzene in the present simulations is
somewhat similar to the trajectories of benzene
derived from the previous molecule dynamics simula-
tion performed by Sastre et al. [12].
Fig. 4(c) depicts the mass cloud of benzene with
interaction energy ranging from 2100 to
220 kJ mol21 (the benzene molecules with interac-
tion energy lower than the most probable energy). It
can be observed that the benzene molecules with rela-
tively lower interaction energy are almost adsorbed
close to the S1 site. Fig. 4(b) shows the mass cloud
ranging from 2100 to 218 kJ mol21; besides benzene
adsorbed close to the S1 site, some benzene molecules
near the S2 site also fall into this energy interval.
Comparing three mass clouds in Fig. 4, it is obvious
that the interaction energy of benzene at the S3 and S4
sites is mainly higher than 218 kJ mol21.
The interaction energy of benzene adsorbed close
to the S1 site is only about 220 kJ mol21, but these
molecules seem to move within a restricted area in the
10-MR channels. Although there is a high degree
of tortuosity in the circular channels, the circular
10-MR channels in ITQ-1 are so small (only
4.0 AÊ £ 5.5 AÊ ) that it becomes dif®cult for benzene
to diffuse through the 10-MR sinusoidal channels
of ITQ-1. Our results on the S1 site occupancy are
in good agreement with the recent molecular
dynamics simulations of the diffusion of benzene
and propylene in ITQ-1 [12], which indicates that
the benzene molecules move within a restricted
area around the minimum energy position from
the trajectories of molecular dynamics.
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±23 15
Fig. 4. (a) Mass cloud for benzene in the ITQ-1 cavities. (b) Mass cloud for benzene with interaction energy ranging from 2100 to
218 kJ mol21. (c) Mass cloud for benzene with interaction energy ranging from 2100 to 220 kJ mol21.
For the product of the alkylation of benzene
with propylene, it will be more dif®cult for the
larger cumene to migrate through 10-MR channel
systems. The previous results of Perego's study
compared the diffusion of cumene in MCM-22
and ZSM-5, where the activation energies had
been calculated to be 90.0 and 18.6 kcal mol21
in the sinusoidal channels of the MWW structure
and in the channels of ZSM-5, respectively. The
combination of the present work and the previous
simulations suggests that the benzene and the
larger cumene molecules will ®nd it very dif®cult
to penetrate through the 10-MR sinusoidal
channels.
At the S2 site, the benzene molecules also
possess relatively lower interaction energy,
although a little higher than those at the S1 site,
which distribute in a generally localized manner.
At the S2 site, the plane of benzene is observed to
prefer being parallel with benzene near the S1
site. We assumed that the benzene molecules at
the S2 and S3 sites would produce relatively
strong aromatic stacking interactions.
The other two interesting sites are located near
the center of the 12-MR supercages (the S3 and
S4 sites). In our model, the interaction energy of
benzene adsorbed close to those two sites is
higher than 218 kcal mol21. It is apparent from
Fig. 4(a) that the benzene molecules near the S3
and S4 sites are considerably delocalized in the
vicinity of their preferred sites of adsorption,
which are quite different from those near the S1
and S2 sites.
Each supercage is connected to six other super-
cages through 10-MR windows, and therefore,
intercage motion is, in principle, possible. But
partly due to the size and the position of the 10-
MR interconnecting windows, the benzene molecules
near them should be energetically unfavorable,
and relatively high activation energy must be
needed for benzene molecules to migrate from
one supercage to another through 10-MR
windows. In Fig. 4(a), benzene is not observed
in the 10-MR interconnecting region, so it can
be concluded that the migration of benzene mainly
happens in the same supercage, and the intercage
motion is somewhat dif®cult from the viewpoint
of interaction energy. Anyway, the migration of
benzene along 12-MR cavities is much easier
than that along 10-MR channels.
3.3. The diffusions of propylene in the two
independent channel systems
The evolution of the loading and energy of
benzene shows that propylene achieves the equili-
brium faster than benzene. At 300 K and 1 atm,
the full loading of propylene is about 83.7 mole-
cules per 8 ITQ-1 unit cells. The mean total inter-
action energy between the sorbates and the zeolite
is nearly 29.41 £ 102 kcal mol21. Fig. 3(b) depicts
the energy distribution of interaction energy
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±2316
Fig. 5. Mass cloud for benzene in the ITQ-1 cavities along the xy orientation.
between propylene and ITQ-1 zeolite at full
loading. The energy distribution of propylene is
quite similar to that of benzene, roughly single-
peaked and an obvious shoulder. But the interac-
tion energy for propylene is much higher than that
for benzene, which means propylene may be more
unfavorable in zeolite than benzene; moreover, the
shoulder for propylene is not obvious.
The mass cloud of propylene indicates a completely
different picture compared with that of benzene (Fig.
5). From Fig. 6, it can be seen that compared
with benzene, the propylene molecules located in
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±23 17
Fig. 6. (a) Mass cloud for propylene in the ITQ-1 cavities. (b) Mass cloud for benzene with interaction energy ranging from 2100 to
210.5 kJ mol21. (c) Mass cloud for benzene with interaction energy ranging from 2100 to 212 kJ mol21.
Fig. 7. Mass cloud for propylene in the ITQ-1 cavities along the xy orientation.
the 10-MR channels are generally localized, but a
much wider area is covered by the smaller propylene
molecules. Careful inspection of Fig. 6 shows that the
propylene molecules are not observed in the channel
intersections, indicating that the channel intersections
correspond to the locations of higher energy.
Although propylene seems to move more freely in
10-MR channels compared to benzene, a certain acti-
vation energy seems to be needed for propylene to
migrate through the 10-MR channels freely.
A completely different picture can be observed for
propylene (Fig. 7) in the 12-MR supercages compared
with benzene, where the propylene molecules almost
®ll all the possible positions in one supercage. It
should be noted that the propylene molecules can be
steadily located in the short 10-MR conducts around
3 AÊ long, very different from the mass cloud of
benzene. Fig. 6(b) shows the mass cloud of propylene
with interaction energy ranging from 2100 to
212 kJ mol21 (the sorbate molecules with interaction
energy lower than the most probable energy), and it
can be observed that the propylene molecules within
that energetic interval are mainly distributed in the
interconnecting area of 10-MR channels and the
short 10-MR conducts. Obviously, the propylene
molecules located in some areas between two adjacent
12-MR supercages are energetically favorable, which
means that propylene can cross from one 12-MR
supercage to another easily and does not even require
any activation energy.
Considering the localization and adsorption of
propylene and benzene in two separate channel
systems, it can be concluded that the reaction
will mainly occur in supercages, not in 10-MR
channels, because the benzene molecules and the
larger product of alkylation are very dif®cult to
migrate through 10-MR channels. Moreover, it
can be concluded that the alkylation on MCM-22
will be partly diffusion-controlled, and the diffu-
sion of benzene in 12-MR supercages and espe-
cially the out-diffusion of cumene will be the
controlling steps of the alkylation reaction.
3.4. Predictions of adsorption isotherms for benzene
and propylene
In order to investigate the adsorption behavior of
benzene and propylene thoroughly, a series of simula-
tions has been performed to get the adsorption
isotherms. In order to be compared with some experi-
mental results, the temperature of the simulations is
set at 315 K, and the pressure ranges from 0.0 to
3.0 kPa. The calculated adsorption isotherms of pure
propylene and benzene in ITQ-1 at 315 K are shown
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±2318
Fig. 8. Simulated adsorption isotherms of benzene and propylene at 315 K and experimental values for toluene.
in Fig. 8. Because the experimental isotherms for
propylene and benzene are not available, the experi-
mental isotherm of toluene is used for comparison
[22]. Certainly, the isotherms of benzene and toluene
cannot be quantitatively compared, but their struc-
tures are similar and their interactions with the zeolite
lattice are only slightly different, so their adsorption
behavior in the zeolite lattice should be similar to
some extent. Only from the viewpoint of volume,
benzene is relatively small, so its loadings should be
remarkably higher than those of toluene, which can be
well deduced from our simulations (Fig. 8). The
experiments of infrared spectroscopy and adsorp-
tion±microcalorimetric studies have been applied to
the adsorption±diffusion behavior of toluene, meta-
and ortho-xylene, and 1,2,4-trimethylbenzene with
different kinetic diameters in MCM-22 [22], and the
adsorption isotherms have validated that the zeolite
uptake signi®cantly relies on the size of the adsorbate
molecules. It can be observed that the uptake of m-
xylene is about half the value of toluene. The value of
o-xylene is much lower and approximately the same
as that of 1,2,4-trimethylbenzene. From our simula-
tions, it can be noted that the uptake of benzene is
much higher than the value of toluene, which also
accords with previous research [22].
The most obvious difference between the adsorp-
tion isotherms of benzene and propylene is that the
loadings of propylene are signi®cantly larger than
those of benzene at low pressure, which seems not
to agree with the laws derived from the isotherms of
some aromatic compounds. For aromatic compounds,
their structures do not exhibit signi®cant difference,
and interactions between sorbate/zeolite and sorbate/
sorbate should be similar, so the adsorbed amounts are
mainly concerned with the size of the sorbate mole-
cules. In the case of benzene and propylene, their
structures are quite different; besides the geometric
factor, the energetic factor will also contribute a lot
to their loadings. The comparison of benzene and
propylene reveals that the interaction energy between
benzene and the zeolite framework is obviously lower
than that between propylene and the zeolite frame-
work. In previous simulations at 350 K and 1 atm, it
has been validated that the interaction energy for
propylene is much higher than that of benzene,
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±23 19
Fig. 9. Two energy minimum conformers of benzene near the abcd plane in the 10-MR window. The benzene molecule is represented with a
ball-and-stick model.
which means that in zeolite cavities propylene may be
more unfavorable than benzene. So in some condi-
tions, especially under low pressure, the unfavorable
energy will make the uptake of propylene lower than
that of benzene.
3.5. Migration of benzene through the interconnecting
10-MR windows
Previous calculations have hypothesized that the
reaction will mainly happen in 12-MR supercages
and the alkylation process will be partly affected by
diffusion of reactants especially benzene and its
products. Obviously, the intercage motion of benzene
will be the most important factor to be considered. In
order to get a clearer picture about the migration of
benzene from one supercage to another, a single
benzene molecule has been observed following a
very simple path connecting two 12-MR supercages.
During the calculations, the zeolite structure is rigid,
but the conformational ¯exibility of benzene is
allowed to be considered, so the potentials include
three terms: Lennard-Jones and Coulomb potential
between zeolite and benzene, plus the internal poten-
tial of benzene.
Energy minimizations under constraints were used
for the systematic search for local minima. First,
considering that during the migration process, the
potential barrier may be near the 10-MR interconnect-
ing areas, the benzene molecule was placed near the
10-MR window as the starting point of its path. Atoms
a±d (Fig. 9) are four Si atoms in the 10-MR window,
which construct a plane. First, the gravity centers of
the plane and the benzene molecule were superim-
posed. Near the starting point, the benzene molecule
was rotated systematically and the energy minimiza-
tions were performed to ®nd the potential minima.
The results indicate that near this point, two potential
minima for benzene can be found (Fig. 10). Moreover,
from the viewpoint of energy, the ®rst conformer is
preferred. Second, we forced the benzene molecule to
translate through the supercage along the axis of
symmetry, and a simple energy minimization strategy
was applied to optimize the conformation of the
benzene molecule to ®nd the local potential minima.
Third, we divided the corresponding path into small
steps 0.25 AÊ distant from each other, and for all the
corresponding conformations, we calculated the inter-
action energy between the benzene molecules and the
zeolite framework. Fig. 11 shows the minimum
guest±host interaction energy as a function of the
distance between the center-of-mass of benzene and
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±2320
Fig. 10. Energy pro®le for benzene in ITQ-1 as a function of the distance between the center-of-mass of the molecule and the center of the abcd
plane while it is pulled from cage to cage. Points A±C correspond to the potential maximum position, and points D±F correspond to the
potential minimum position.
that of the plane abcd. Because of the symmetry of the
zeolite, the energy function is mirror symmetric with
respect to the plane abcd. Only three crystallographi-
cally different types of energetic minima are
observed.
The minima at 0 AÊ corresponds to the starting
point, while 3.75 and 10.25 AÊ seem to correspond to
the S4 adsorption site inside the supercages. Mean-
while, there exist three potential barriers along the
pathway: A, B, and C (Fig. 11), among which the
potentials at A and C are much higher than that at
B. The energy pro®le of benzene as it is pulled from
cage to cage in the ITQ-1 structure is similar to the
previous simulations by Sastre et al. [12]. Obviously,
the potential barriers should control the diffusion of
benzene through 10-MR windows. From Fig. 11, it
can be seen that A and C points are all near 10-MR
windows, and they are symmetric to plane abcd.
Moreover, the separate contributions to the activation
energy have been calculated (Fig. 10). The ®rst contri-
bution comes from the Lennard-Jones energy, and the
second one comes from the electrostatic interaction
between the benzene molecule and the zeolite frame-
work plus the conformational energy of benzene. As
in Fig. 10, the Lennard-Jones potential obviously
contributes more than the electrostatic potential (the
internal conformational energy of benzene can be
ignored). The energy necessary to cross from cage
to cage is around 24 kcal mol21. Previous calculation
results from molecular dynamics [12] have predicted
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±23 21
Fig. 11. Top view of the ITQ-1 structure showing the potential barriers of benzene if it is pulled from cage to cage. Points A±C correspond to
the potential maximum position.
Fig. 12. Side view of the ITQ-1 structure showing the critical points
(A±F) in the diffusion path followed by the benzene molecule as it
is pulled from cage to cage. Points A±C correspond to the potential
minimum position, and points D±F correspond to the potential
maximum position.
that the values are ranging from 15 to 20 kcal mol21,
which generally agrees with our predictions. The
deviation mainly comes from the different potential
parameters and different partial charges adopted in
this paper.
Along the energy pro®le, there exists another
potential barrier at position B, which is just in the
center of the 12-MR supercage. If we treat the lowest
energy minimum in the benzene path as the adsorption
S4 site, therefore, one benzene molecule that migrates
from one S4 site to another in the same 12-MR superc-
age will only need 6 kcal mol21, which is much lower
than that for intercage migration. That is to say, the
migration of benzene mainly happens in the same
supercage and the intercage motions should need
certain activation energy.
Certainly, the diffusion path of the benzene mole-
cule is very simple and ideal (Fig. 12), and it only
corresponds to very low loadings of benzene. In
normal conditions, especially for high loadings, the
presence of the interaction energy of the sorbates
will in¯uence the activation energy greatly, but the
present study can also afford some useful information
about the diffusion behavior. Because certain large
activation energy is needed for benzene molecules
to cross from one cage to another, increasing the
temperature would possibly increase the probability
of observing benzene intercage motions.
4. Conclusions
GCMC simulations have been performed to simu-
late the location and adsorption of benzene and propy-
lene in a purely siliceous MWW structure (ITQ-1).
Two separate simulation processes have been carried
out to explore the locations and possible adsorption
sites in each channel system of the ITQ-1 at 300 K and
1 atm. From calculation results of different distorted
models, it can be concluded that the minor change of
the zeolite framework does not introduce noticeable
effects on the simulation results. The mass clouds
from GCMC simulations indicate that benzene and
propylene show high localization in 10-MR channels,
which is due to the small size of the 10-MR openings
of the sinusoidal systems. In 12-MR supercages, the
benzene and propylene molecules show quite different
mobilities. The spatial distribution of benzene in a
12-MR supercage can be clearly divided into three
sites of adsorption: one site near the 10-MR facing
the 6-MR in supercage and the other two near the
center of the 12-MR supercages. So we can conclude
that the migration of benzene molecules mainly
happens in the same supercage, and the intercage
motions should need certain activation energy. In
the case of propylene, the sorbate molecules not
only almost ®ll all the possible positions in one
supercage, but also can be steadily located in the
short 10-MR conducts around 3 AÊ long, which
means that propylene can cross from one supercage
to another very easily and does not require obvious
activation energy. The adsorption isotherms of
benzene and propylene at 315 K and 0±3.0 kPa
have been predicted, and the adsorption isotherm
of benzene coincides with the trend from experi-
ments on a series of aromatic compounds. At low
pressure, the loadings of propylene are signi®-
cantly lower than those of benzene, which are
mainly caused by the relatively unfavorable inter-
action energy between propylene and the zeolite
framework.
From the calculations for the activation energy of a
benzene molecule in the intercage process through 10-
MR windows, it can be found that at the simulated
conditions the thermal energy allows the mobility of
benzene inside the cavity but not from one 12-MR
cavity to another. By analyzing the locations of
benzene and propylene in the ITQ-1 framework, the
alkylation of benzene and propylene will mainly
happen in 12-MR supercages at the external surface
or close to the external surface.
Acknowledgements
This project is supported by NCSF 29992590-2 and
29573095.
References
[1] P.R. Pujado, J.R. Salazar, C.V. Berger, Hydrocarbon Process.
55 (1976) 91.
[2] H. Miki, US Patent, 4347393, 1982.
[3] E.K. Jones, D.D. Dettner, US Patent, 2860, 1958.
[4] A. Wood, Chem. Week 17 (1994) 34.
[5] W.W. Kaeding, R.E. Holland, J. Catal. 19 (1988) 212.
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±2322
[6] J. Chen, Wordwide Solid Acid Process Conference, Houston,
TX, November 14, 1993.
[7] K.H. Chandavar, S.G. Hedge, S.B. Kulkarni, P. Ratnasamy, J.
Chem. Technol. Biotechnol. 34A (1984) 165.
[8] J.C. Cheng, T.F. Degnan, J.S. Beck, Y.Y. Huang, M. Kalya-
naraman, J.A. Kowalski, C.A. Loehr, D.N. Mazzone, Stud.
Surf. Sci. Catal. 121 (1999) 53.
[9] M.E. Leonowicz, J.A. Lawton, S.L. Lawton, M.K. Rubin,
Science 24 (1994) 1910.
[10] M.K. Rubin, P. Chu, US Patent, 4954325, 1990.
[11] C. Perego, S. Amarilli, R. Millini, G. Bellusi, G. Girotti, G.
Terzoni, Microporous Mater. 6 (1996) 395.
[12] G. Sastre, C.R.A. Catlow, A. Corma, J. Phys. Chem. 103
(1999) 5187.
[13] M.A. Camblor, A. Corma, M.J. Diaz-Cabanas, C. Baerlocher,
J. Phys. Chem. 102 (1998) 44.
[14] E.D.V. Burchart, PhD thesis, 1992.
[15] S.L. Mayo, B.D. Olafson, W.A. Goddard, J. Phys. Chem. 94
(1990) 8897.
[16] Cerius2 Simulation Tools User Reference, Molecular Simu-
lation Inc., San Diego, CA, 1998.
[17] Mopac 7.0 User Guide, Quantum Chemistry Program
Exchange (QCPE), Indiana University, USA, 1993.
[18] Q.S. Randall, T.B. Alexis, N.T. Doros, J. Phys. Chem. 97
(1993) 13742.
[19] P.A. Van Tassel, H.T. Davis, A.V. Mccormick, J. Chem. Phys.
98 (1993) 4173.
[20] M.A. Camblor, A. Corma, C. Baerlocher, J. Phys. Chem. 102
(1998) 44.
[21] M. Czjzek, H. Fuess, T. Vogt, J. Phys. Chem. 95 (1991) 5255.
[22] A. Coma, A. Corell, J. PeÂrez-Pariente, J.M. Guil, R. Guil-
LoÂpez, S. Nicolopoulus, J. Gonzalez-Calbet, M. Vallet-Regi,
Zeolite 17 (1996) 7.
T.J. Hou et al. / Journal of Molecular Structure (Theochem) 535 (2001) 9±23 23